View Issue Details

IDProjectCategoryView StatusLast Update
0003299OpenFOAMBugpublic2019-06-28 12:27
Reporterpantelis Assigned To 
PrioritynormalSeverityminorReproducibilityalways
Status newResolutionopen 
PlatformLinuxOSCentosOS Version6.5
Summary0003299: Incorrect particle-wall interaction in OpenFOAM 6
DescriptionI am working on particle filtering in a bend pipe geometry. I recently switch from version OF2.2x to OF6. I found out that the interaction of particles with the walls in not correct in OF6. I observed that a particle sticks on the wall only when it’s center touches on the wall. However the particle must stick on the wall when it’s surface touches the wall. In other words, it’s diameter should be taken into account to check whether there is contact to the wall. I have tested this in OF2.2x and it was working properly. I attach two relevant images.
I observe the same behavior in OF6 compiled on Centos with Intel compiler and on macOS using Docker for Mac.
I suspect something doesn't work properly in p.hitFace(s, cloud, ttd); member function (line 361 in /src/lagrangian/intermediate/parcels/Templates/KinematicParcel/KinematicParcel.C)
Steps To ReproduceFind the case files to reproduce the case in this dropbox link:
https://www.dropbox.com/sh/budrftdp2smq9ld/AADiGiY0LZOw8iuYgs6azFbGa?dl=0
Run icoUncoupledKinematicParcelFoam from 0-5s.
Under a uniform velocity field, a particle with diameter 0.003088m is injected at the inlet of the bent, very near to the upper wall.
The particle is deposited on the wall at x=2.28156. It should be deposited a bit upstream when it's diameter is taken into account (x=2.19506)
TagsNo tags attached.

Activities

pantelis

2019-06-28 08:53

reporter  

OF6.png (143,220 bytes)   
OF6.png (143,220 bytes)   
OF22x.png (109,482 bytes)   
OF22x.png (109,482 bytes)   

will

2019-06-28 11:12

manager   ~0010535

The behaviour in OF 4 and earlier was to search the current cell for wall-interactions at every tracking step. This was inefficient for the vast majority of usage cases that do not need particles to interact with boundaries at their radius, and it was unreliable in that if a particle was positioned such that it's radius should intersect a patch face not in the current cell that interaction would be ignored. For these reasons the functionality was removed and (most) particles from version 5 onwards are considered as point masses with respect to patch interaction.

The only Lagrangian element that retains the ability to interact with a wall at it's radius is the colliding parcel. As such, the only way of representing collisions in this way at present is to use a colliding-cloud-type solver (which you are, icoUncoupledKinematicParcelFoam) and switch on the pairCollision model. This is expensive to do as it will calculate all the particle-particle interactions, too.

We have some funding for, and are working on, splitting up the pair collision model so that the particle-wall collisions can be used independently from the the particle-particle collisions. That would allow you to model these collisions reliably and efficiently, though it is still likely to be more expensive than before. Watch this space.

pantelis

2019-06-28 12:27

reporter   ~0010536

Thanks for the clarification Will.

Issue History

Date Modified Username Field Change
2019-06-28 08:53 pantelis New Issue
2019-06-28 08:53 pantelis File Added: OF6.png
2019-06-28 08:53 pantelis File Added: OF22x.png
2019-06-28 11:12 will Note Added: 0010535
2019-06-28 12:27 pantelis Note Added: 0010536